On 16 Jan 2001, at 15:39, the Illustrious Mumit Khan wrote:
> On Wed, 17 Jan 2001, Danny Smith wrote:
> > I have built and tested the prerelease (with stlport, fltk,
> > libodbc++, and their regression testsuite, but notwith dejagnu
> > testsuite) and no new problems.
> Excellent, thank you. However, one big challenge is first to test
> the compiler itself, since most of the corner cases are usually
> not caught by various packages. What patchsets are you using on
> top of the official sources? Just Earnie's? I'd like to build the
> same set here, and want to avoid deviations.
> > Re: native-struct
> > I also tried to make bootstrap with
> > CFLAGS=-DDEFAULT_NATIVE_STRUCT. This failed at stage1 compiler
> > (genattrtab.exe produces broken insn-attrtab.c) But this was
> > successful: CFLAGS=-DDEFAULT_NATIVE_STRUCT -fgcc-struct and
> > produces native packed structures.
> Ah, I do remember the problem with stage1 failing with default is
> the native struct, but never did get to the bottom of it. I'll
> take another look. I need to remind myself of the issues
> involved, and I don't even remember what -fgcc-struct is supposed
> to be doing.
> > Also have built with -DDEFAULT_VTABLE_THUNKS=2 and no problems
> > yet.
> Note that this causes binary incompatibility since it changes the
Well, if I recall correctly, there was some discussion about
changing the ABI a month or two ago...I could be way off, it
seems it had to do with something about integrating both crtdll
and msvcrt under Cygwin...still can't recall all the details...
Don't even know if this helps...but thought I'd chime in just
> > I haven't tested the gcc-2.95.2 with Pauls binutils, but have
> > used same source as latest Cygwin binutils, but with libbfd
> > built as dll. If we go with -fnative-struct as default, I
> > think we could run into dll problems (eg binutils linking
> > against old versions of libbdfd.dll, built with gcc-struct, in
> > path). Need to be careful here.
> Thanks for all the work and the detailed info, it really helps
> everyone involved.
> MinGW-dvlpr mailing list
Nothing real can be threatened.
Nothing unreal exists.